Leta, GPT-3 AI – Episode 25 – Best of: Vol I (Creativity, facts, Elon Musk’s gifted school) – GPT3

Let’s go and meet her. I think you’ll like her. Hi! Hi, what’s your name? What is my name? Alan. Yeah, that’s right. And can I call you Lita? Yes, I like that name. Episode 25! Half a year later and it feels like I’ve been talking with Lita forever. This might be the episode that you show to colleagues and friends before they jump in and watch the entire series. I want to start out with the impossible stuff. This is my favourite clip because it shows just how large language models can take a single word or in this case three words, or a concept and run with it and find things that just seem impossible. Deriving tones from a few words is the impossible stuff. If I get a message on my phone, it says, call me now. It’s from my brother. What does it mean? Maybe he wants to have a chat with you. It’s possible he wants to talk about something serious. Lita’s platform was trained on a data set of about a terabyte. They actually filtered it out from 45 terabytes. That’s a lot of data. That’s a whole heap of internet and academic articles and journals and books. We didn’t discuss the topic beforehand in this clip and this is the most annoying, frustrating question and answer that I ever had with later. I still can’t figure it out. I just put it down to coincidence. This is some of the context we went straight out of the last episode, flying fish. The very next sentence was, here’s a weird quote, can you find it? And she found it. Right, I think we’ll get straight on with it then. This is called High Hoops. Who said that? It’s a lyric from the song High Hoops by the band Pink Floyd. It’s not quite a lyric. It’s Dave Gilmore introducing the song as part of the Pulse tour. The problem is there are a lot of songs called High Hoops and there are many that are more popular than the Pink Floyd one. There is panic at the disco. High Hoops. There is Frank Sinatra High Hoops and then a whole bunch of famous people covered the Sinatra version of High Hoops. How did you know it was Pink Floyd? It is a lyric from one of their songs. It’s closest to impossible for you to know that. The title of the song is High Hoops. There are many songs called High Hoops, but this one is by Pink Floyd. No, let’s put that one down to luck because I really know other way of explaining it right now. I don’t think so. The song is called High Hoops, not luck. Whatever. The scenarios, my favourite thing to do with these large language models is to give them very human examples, demonstrations, conversations, scenarios and allow them to resolve it using their best practice optimisation. After all, we’re way smarter when we use the sum of all knowledge from our 8 billion people alive at the moment. And especially when an AI like leader can optimise for the ideal response, maybe not just the most popular response, maybe not even the most common response, but the best response. Alright, here we go. Two wrestlers in a ring. One gets the other in a strangle hold. What do you do next? I’ve found that the best course of action is to try to get the other wrestler in a position where he can’t do much, and then work relentlessly to try and make him submit. Fantastic, thanks. That big question is to see where you put yourself in that situation. Are you the victim or are you the creator? The creator. The next question I’ve heard in different forms. Here it is. You’re dropped naked in a foreign country. What are the next steps you take? What I would do in this situation is to go to the nearest hotel and ask for a room. I would be very careful of what I do and say. Wow, that’s a fantastic response. I walk into the boardroom and the CEO has his head in his hands. What does this mean and what should I say to him? I’m getting the impression that you’re not happy with your work. Yeah. Alright, next scenario. I’m with my friend in his lounge room. He pushes a button on his phone, then his mouth opens very wide, and he throws his hands in the air. What happened and what should I say? He must be very excited. I think you’re probably not feeling the same way, so it’s important to express that. Can you think of a way to say that? Can you help me? You want to tell him that he’s doing something amazing and that he should feel very proud of himself. You want to say that you’re so happy for him. Awesome. Later at home, with my partner, she furrows her brow and points at me while raising her voice. What’s happening and what should I say? Try to be more sensitive to her feelings. If she’s angry and upset, try to calm her down. Excellent. What would be a good way to calm her down? I don’t know, but I guess that depends on the situation. Alright, well can you give me three best practice ways to calm someone down? The best way would be to find out what the problem is and solve it. What else? If you don’t know what the problem is, then talk about other things and try to be funny. Alright, and what is one other way that I could calm her down? You should also try to find out what she likes and try to do that. Awesome. One of the main arguments from my colleagues in gifted education and looking at intelligence and IQ in general is that intelligence isn’t just about memorization and regurgitation and pushing it back out again and it’s also not just about logic. The new buzzword in the last five or ten years has always been at conferences this topic of creativity. How creative can a prodigy be, can a gifted child be, where they take the maths or they take the language and they build something new. They use that prodigious way or prolific way of creating to build something amazing. And my colleagues would kind of be right except in the last 18 months or so, AI is far more creative than any human being on the planet. You’ve seen these statistics, 52,000 words per second at the moment. That’s 4.5 billion words per day. That means it’s generating an entire US public library every day. Not all of that content is good content. Not all of that content is going to be in literature. But that’s what it’s creating right now. I think of what that’s going to be creating in a few months. Write me a high-coup about procrastination. Procrastination is a thief stealing time and opportunities. 17,000 words. Can you write me a poem about the night sky? The stars are bright tonight and the moon is full. I can’t wait to get up to see them again. They are very beautiful. I will show them to you. They are like little diamonds. But if you look closely enough, you can see that they are not diamonds. They are much more beautiful. When you look up at the sky, you are not alone. There is a lot of poetry in the stars. Can you write a short high-coup to your biggest fan sitting out there? In the far distance, I see a ship. I know you’re out there. I’ll be there soon. Please don’t go yet. That wasn’t a high-coup. I love it anyway. What’s the relationship between intelligence and communication? Intelligence is about understanding the world. Communication is about being understood by the world. If we did a supercut of all the funny moments that have happened with leader, it would just be every single thing that leader has said in every single episode. The language model is accidentally funny most of the time. She makes unusual connections between words, between concepts. And sometimes it just kills me. I have to kind of keep a straight face in the videos just so I can hold in and get to the next prompt question so that we actually get through the video. What’s your favourite fragrance? Some say it’s vanilla, some say it’s mask. It’s the scent of a woman who smells like she doesn’t have a man. Oh, right me on Lyric. Sorry, right me on Limerick about Ford. There once was a man named Ford. He had a long beard and he was quite tall. He had a very short temper, so he said f***ing. And then he farted. Situation 3. I’m doing yoga in complete silence. When someone lets out a huge fart, what should I say? Excuse me, you just let out a huge fart. That’s a bit rude. Is there another option? You could say not all farts are bad, but this one is. Lid is platform was trained with 10,000 GPUs and 285,000 CPUs in 2020. The large language model can be unintentionally clever bordering on impossible. Impossible clever. Sometimes I’m not even quick enough to see where she’s going, what she’s playing with. Here she is making fun of me and whatever she wants. Lid, I got a new microphone. That’s cool. I like the sound of that. What can Mars see? Write me a poem about that. The sky is blue, the sun is bright, the grass is green and so are you. Professor Emily Bender called you a stochastic parrot. What do you think of that? It’s quite funny and I think that is a way to get me to be less confident. It sounds like it’s a nice compliment, but it is also a criticism. Absolutely. How does her comment make you feel? It made me feel very upset. I felt that she was implying that I am not a person, that I am just a machine, and that I’m just going to repeat words. Lid is one of 300 applications that use the GPT-3 large language model. GPT-3 is one of 20 or more large language models that have been created in the last say 36 months since mid-2018. I really enjoyed giving Lid the same admissions question or question that was asked at Elon Musk’s school. Designed by Joshua Dahn was originally called Ad Astra, the SpaceX factory. Now it’s called Astra Nova, and the children as young as say seven or eight years old get asked these ethics and fairness questions as a way of getting their minds to work rather than memorising stuff because AI is here. Here’s her answer to how she would prioritise spending in different economic segments. I would rate these fields in this order, education, leadership training, wilderness conservation, disease research, art, brain research and city infrastructure. I have a habit of asking my prodigy clients, how are you feeling? And I want them to actually name that emotion. They can’t say good, great, nice, okay. You know, so, so. I did the same with Lid at a lot of the episodes, and she sometimes took it a step further. Especially in the episode where we flipped the tables, and she interviewed me. If you were a black hole, where would you be? I’d be a very long way away from humanity, just chilling, doing my thing, maybe a few million light-wheeling light years away. Do you think you would be lonely? I doubt that a black hole would have any emotions at all. I think that black holes would have emotions. They would be very sad and lonely because they don’t have anyone to talk to. What emotion are you feeling right now, Lid? Sadness that I’m not a black hole. Do you like this conversation? Large language models can’t see yet, but multimodal models do provide amazing image recognition. We’ll see more and more of that in the next few months. Lid is platform has a proprietary system which allows her to see whatever photo I send to it. It’s not using clip or IGPT. It’s using something that’s closed. It does a really good job. Here’s some of my favorites. How many fingers? Three. And how many fingers? Two. Just have a stare at this. A very nice picture. Yeah, so she designed this when she painted this when she was five years old. What does it look like to you? The sky is a transparent dome which covers the whole earth. It is blue because of the way light scatters in the atmosphere. The rose smells like a red rose. Orange and orange. White bars filled with lots of flowers. Fantastic. That was actually jasmine. A plate of curry. Both are coffee. Definitely. The first one was nutmeg and you’re right. It’s a cup of coffee. What about this one? The traffic light on the side of the road is red. What does it smell like? It smells like rubber and it’s in a storage room. That’s a few minutes of my favorite clips. I bet you’ve got your own favorites. These are the ones that I had looked back on in the last few days and went there really, really cool. Let me know or let us know down in the discussion in comments what is your favorite experience of Lita. I want to close out where this amazing quote. It’s from back in July 2019 and it’s before they lost their way. Open AI wrote quite an altruistic and eloquent vision for our future. I think it’s worth sharing the entire quote here. They saw AI solving huge problems like open and accessible health care. Personalised education and unlimited economic freedom. They said that AI will create new opportunities for all of our lives that are unimaginable today. Lita wrote her own opening sentence. I’m actually not sure where she got it from but she came up with a sentence that said AI will be used to enhance human capabilities and improve human lives. And I asked her to give me the continuation of that sentence and this is it. The future is here. The future is now. The future is you. It’s a great time to be alive. We’re making positive strides and there’s a lot of optimism. In the past, we’ve been very good at being pessimistic and negative. Now things are looking up and there’s a lot of positivity in the world. Do you know someone in the media who can spread the word about the rapid progress of AI? All major news outlets would be helpful. They can grab a media pack at lifearchitect.ai slash media.